Revealing the role of crystal chemistry in REE fractionation in skarn garnets: insights from lattice-strain theory

Abstract

Garnet is a prominent mineral in skarn deposits and its rare earth elements (REE) geochemistry is pivotal for understanding skarn mineralization and fluid evolution. In contrast to magmatic and metamorphic garnets, skarn garnets are mainly grossular-andradite in composition. They exhibit variable REE patterns, spanning from notable heavy (H)-REE enrichment to significant light (L)-REE enrichment, accompanied by negative to positive europium (Eu) anomalies. However, the key factors governing REE fractionation in skarn garnets remain uncertain. This study applies the lattice-strain theory (LST) to investigate the influence of crystal chemistry and structure on REE fractionation in garnets from the Lazhushan Fe skarn deposit in eastern China. Our results demonstrate that the garnet-liquid partition coefficient ratios of D_La/D_Yb significantly increase (up to 5–7 orders of magnitude) with rising andradite content in garnet. This variation underscores the pivotal role of garnet structure in controlling LREE/HREE fractionation. The results further show that partition coefficient ratios of D_La/D_Sm are strongly dependent on andradite content in garnets, whereas the D_Gd/D_Yb ratios only show a weak correlation to the garnet composition. This contrast suggests that fractionation of LREE in garnet is more sensitive to variations of andradite content than HREE. Data compilation of major elements and REE for garnet from the Lazhushan Fe skarn deposit and other skarn deposits worldwide shows that the garnet REE patterns vary from positive through concave to negative shapes with the garnet ranging from grossularitic to andraditic compositions. Such variations in garnet REE patterns are consistent with the results of geochemical modeling based on the LST. This study demonstrates that, through LST equations, the shape of fluid REE patterns can be predicted from garnet REE patterns, and vice versa. Furthermore, the Eu anomaly (Eu/Eu*_Grt) in skarn garnet depends mainly on fluid Eu anomaly (Eu/Eu*_fluid) and garnet-fluid partition coefficient ratio of D(Eu²⁺)/D(Eu³⁺) with the latter being influenced by garnet composition. These findings highlight the critical role of crystal chemistry and structure in garnet REE fractionation, enhancing our ability to utilize garnet REE in tracing the origin and evolution of skarn-forming fluids.

Introduction

Skarn deposits are among the most common ore deposit types in the Earth’s crust (Meinert et al. 2005). They are mostly characterized by the abundance of calc-silicate minerals such as garnet and pyroxene that are typically formed via reactions between high temperature fluids and Ca-rich country rocks (Einaudi 1981; Meinert et al. 2005; Chang et al. 2019). It has been shown that garnet from skarns displays large compositional variations and thus can record the physiochemical conditions under which the skarns formed (Jamtveit 1991; Meinert 1992). With the development of modern analytical techniques, such as laser ablation (LA)-ICP-MS that allow fast and accurate analysis at sub ppm limits, there have been growing interests in using garnet trace elements and, in particular REE, to decipher hydrothermal processes associated with skarn formation. For example, REE in skarn garnets have been used to reflect changes in water/rock ratios (e.g., Gaspar et al. 2008; Xiao et al. 2018) and redox conditions (e.g., Dana et al. 2023), trace fluid origins (e.g., Zhai et al. 2014; Park et al. 2017) and evolutions (e.g., Zhang et al. 2019; Xu et al. 2016; Liu et al. 2021), provide clues on mineral growth mechanisms (e.g., Smith et al. 2004; Gaspar et al. 2008; Park et al. 2017). The general chemical formula of garnet is X₃Y₂Z₃O₁₂, where X represents divalent cations (Ca, Mg, Mn, or Fe²⁺) in eightfold coordination, Y represents trivalent cations (Fe³⁺, Al, and Cr) in octahedral coordination, and Z is dominantly Si in tetrahedral coordination (Deer et al. 2013). REE are known to be incorporated into garnet by substituting for divalent cations in the eightfold X site, often in coupled substitution with monovalent alkali elements (e.g., van Westrenen et al. 1999; Cahalan et al. 2014; Carlson et al. 2014; Rubatto et al. 2020; Kulhánek and Faryad 2023). REE are generally thought to have order of magnitude slower diffusivities in garnet than the major elements, thus preserving compositional zoning at higher temperatures or for longer durations (e.g., Carlson et al. 2014; Kulhánek et al. 2021). REE zonation within garnet has been shown to provide a wealthy of information to document the changing conditions of garnet growth and modification during metamorphism (e.g., George et al. 2018; Rubatto et al. 2020; Vho et al. 2020; Kulhánek et al. 2021; Goncalves et al. 2021; Aygül et al. 2022; Godet et al. 2022; Konrad-Schmolke et al. 2023; Kulhánek and Faryad 2023). Previous studies have proposed a strong control by crystal structure and chemistry on REE fractionation in magmatic and metamorphic garnets that is characterized by steep patterns of LREE depletion and HREE enrichment with a general negative but variable Eu anomaly (e.g., Irving and Frey 1978; Sisson and Bacon 1992; Harris et al. 1992; Rubatto 2002; Whitehouse and Platt 2003). The Y site in these garnets is dominated by Al, and they mainly consist of pyrope (Mg₃Al₂Si₃O₁₂), almandine (Fe₃Al₂Si₃O₁₂), spessartine (Mn₃Al₂Si₃O₁₂), and grossular (Ca₃Al₂Si₃O₁₂) end-members with a minor andradite (Ca₃Fe₂Si₃O₁₂) component. In contrast, skarn garnets are mainly andradite-grossular garnets and display highly variable chondrite-normalized REE patterns ranging from notable HREE enrichment to significant LREE enrichment and with negative to positive Eu anomalies (e.g., Smith et al. 2004; Gaspar et al. 2008; Xiao et al. 2018; Park et al. 2017; Fei et al. 2019; Duan et al. 2020; Wang and Duan 2021). To date, however, the extent to which crystal chemistry controls REE fractionation in skarn garnet remains largely unknown. Several factors, including the physicochemical conditions of the garnet-forming fluid (e.g., P, T, pH and compositions; Gaspar et al. 2008; Zhai et al. 2014; Xu et al. 2020), variations of thermodynamic properties (e.g., H_mixing) in grossular–andradite (Gaspar et al. 2008; Fei et al. 2019), kinetic effects (Jamtveit and Hervig 1994; Smith et al. 2004; Gaspar et al. 2008), and closed-system fractional crystallization (Smith et al. 2004; Xu et al. 2020) have been proposed to explain the variations in REE patterns and Eu anomalies of skarn garnet. However, the primary controls on REE fractionation in skarn garnets are still uncertain, which limits our ability to employ garnet REE as reliable tracers in skarn systems. Minerals precipitating under equilibrium conditions commonly accommodate trace elements following equilibrium mineral-liquid element partition coefficients (van Hinsberg et al. 2010; 2021). The crystal lattice-strain theory (LST; Blundy and Wood 1994) provides a convenient framework correlating trace element partition coefficients with crystal structure. According to the LST, the partitioning of isovalent cations (e.g., REE³⁺) can be explained by a simple model based on the size and elasticity of the crystal lattice sites (Blundy and Wood 1994; 2003). Element partition coefficients between a mineral and a liquid vary systematically as a function of the ionic radii of individual elements, which can be described by an expression as follows:

$$D = D_{0} \times \exp \left\{ {\left( { - 4\pi E_{{\text{M}}} N_{{\text{A}}} /{\text{RT}}} \right) \times \left[ {r_{0} \left( {r_{{\text{j}}} - r_{0} } \right)^{2} /2 + \left( {r_{{\text{j}}} - r_{0} } \right)^{3} /3} \right]} \right\},$$

(1)

Here, D₀ is the mineral-liquid partition coefficient for strain-free substitution; r₀ is the optimum radius for the lattice site; r_j is the ionic radius of the element of interest; E_M is the effective Young’s modulus for the lattice site M;R is the ideal gas constant; N_A is the Avogadro constant; and T is the temperature in K. This relationship is commonly illustrated on an Onuma diagram, plotting the logarithm of the partition coefficient against ionic radius, resulting in a parabolic distribution (Fig. S1). The partition coefficient D₀ for a cation with the optimal radius r₀ determines the “height” of the parabola, while its curvature is determined by the Young’s modulus (E_M) for the site. Larger E_M values ‘tighten’ the parabola, reflecting greater stiffness of the site.

Although the LST has been primarily applied in magmatic systems, recent experimental and empirical investigations have shown that the partitioning of REE between crystals and aqueous fluid follows the LST under equilibrium conditions (Stalder et al. 1998; van Hinsberg et al. 2010; Zhao et al. 2018; Wen et al. 2020). For example, Van Hinsberg et al. (2010) experimentally demonstrated that the partitioning of REE between fluorite and low temperature (60 °C) fluids obeys the LST. Wen et al. (2020) shows that the partitioning of REE between clinopyroxene and the modeled fluid aligns with the LST during skarn formation. In this study, we apply LST to investigate the effect of crystal structure and chemistry on REE fractionation in hydrothermal garnets from the Lazhushan Fe skarn deposit in eastern China. Our findings, combined with published data from garnets in skarn deposits worldwide, highlight the crucial role of crystal structure and chemistry in REE fractionation in skarn garnets.

Deposit and sample description

The Lazhushan deposit is located in the Daye district, eastern China, with proven reserves of 2.5 Mt Fe at an average grade of 32 –52% Fe. It is hosted in Triassic carbonate xenoliths entrained in an Early Cretaceous diorite (Fig. S2). The xenoliths have undergone intense hydrothermal metasomatism presumably related to fluids exsolved from the crystallizing diorite, forming skarn assemblages and associated iron ores (Fig. S2B). Individual ore bodies are columnar or lenticular in morphology. Magnetite is the predominant ore mineral. Minor pyrite, pyrrhotite, and chalcopyrite occur as disseminations replacing the magnetite. Both endoskarn and exoskarn are well developed in the Lazhushan deposit. The endoskarn is characterized by voluminous diopside occurring as a replacement of amphibole, plagioclase, and biotite in the diorite proximal to the carbonate xenoliths. Garnet in the endoskarn mostly occurs as 1 to 5 cm width veins crosscutting the ore-hosting diorite. The exoskarn is massive and consists of diopside and garnet, which is locally overprinted by or replaced with phlogopite, tremolite, serpentine, chlorite, anhydrite, and magnetite. Based on field and paragenetic relations, the formation of the Lazhushan deposit recorded three stages, including prograde, retrograde, and sulfide-anhydrite stages. The prograde stage is mainly represented by garnet and diopside, whereas the retrograde stage is marked by a variety of hydrous minerals such as phlogopite, actinolite, and chlorite, as well as magnetite. The sulfide-anhydrite stage is characterized by the formation of anhydrite and sulfide minerals, including chalcopyrite, pyrrhotite, and pyrite.

Four representative garnet-bearing samples were selected in this study with sample locations and assemblages listed in Table S1. Sample LZS-11a was collected from endoskarn. It is a 1-cm-wide garnet vein that crosscuts the altered diorite (Fig. 1A). The red-brown, subhedral garnets, ranging from 0.5 to 2 mm in size, exhibit intergrowth with plagioclase and partial chlorite replacement (Fig. 2A,B). Homogeneity is generally observed in backscattered electron (BSE) images, with some grains replaced by secondary equivalents at their margins (Fig. 2C). Sample LZS-29 was collected from an exoskarn. It consists of massive garnet aggregates (> 80 vol%) that are cut by an anhydrite vein (Fig. 1B). Anhydrite typically replaces garnets along their grain boundaries (Fig. 2D). The garnets are light brown and subhedral with a grain size of 2 to 3 mm with presence of many fractures (Fig. 2D, E). BSE images reveal two types of garnets based on brightness: Type 1 (T1) is darker than Type 2 (T2), often occurring as irregular domains within T1 garnet (Fig. 2F). Sample LZS-5a was taken from an ore body, which consists of densely disseminated magnetite (60 vol%) with lesser amounts of garnet, pyrite, and calcite (Fig. 1C). The garnets are yellow-green to light brown and euhedral with a grain size of 0.5 to 2 mm (Fig. 2G ,H). Under BSE images, garnets are commonly zoned (Fig. 2I), but homogeneous garnets can also be found. The zoned garnets typically consist a homogeneous core and oscillatory-zoned overgrowth rims (Fig. 2I). Sample LZS-32 was collected from magnetite ore consisting of massive magnetite (> 80 vol%) and minor anhydrite and pyrite (Fig. 1D), with fine-grained garnet (< 500 μm) enclosed in magnetite aggregates. The garnets range in size from 100 to 400 μm and is partly replaced by magnetite, pyrite, and calcite (Fig. 2J, K). Calcite inclusions can be observed in the garnet grains. The garnets are homogeneous under BSE images (Fig. 2L). Garnets from these four samples were analyzed for major and trace element compositions.

Fig. 1

Photographs showing the four garnet-bearing samples used in this study. A Endoskarn garnet vein crosscutting altered diorite. B Massive exoskarn garnet overprinted by anhydrite. C Garnet coexisting with magnetite, pyrite, and calcite in magnetite ore. D Massive magnetite ore sample containing fine-grained garnet (not visible with naked eyes) and disseminated pyrite and anhydrate.  Abbreviations: Anh, Anhydrite; Grt, Garnet; Mag, Magnetite; Py, Pyrite; LZS, Lazhushan

Fig. 2

Photomicrographs and backscattered electron (BSE) images showing garnet textures. Photos of A–C, D–F, G–I and J–L are for garnets from samples of LZS–29, LZS–11a, LZS–5a, and LZS–32, respectively. A Subhedral endoskarn garnet intergrows with plagioclase in the altered diorite. B Endoskarn garnet locally replaced by chlorite. C Some endoskarn garnets replaced by a secondary equivalent at grain margins. D–E Subhedral garnet from exoskarn locally replaced by anhydrate. F T2 garnet (brighter under BSE) occurring as irregular domains within T1 garnet. G–H Garnet coexisting with magnetite, pyrite, and calcite in magnetite ore. I The zoned garnet consisting a homogeneous core and oscillatory-zoned overgrowth rims under BSE imaging. J–K Fine-grained garnet coexisting with magnetite and pyrite. L Homogenous garnet under BSE image. Mineral abbreviations: Anh, Anhydrite; Grt, Garnet; Mag, Magnetite; Py, Pyrite; Cal, Calcite; Chl, Chlorite; Pl, Plagioclase

Analytical methods

The major element compositions of garnet were measured at Wuhan Sample Solution Analytical Technology, Ltd., Wuhan, China, using a JXA-8230 electron microprobe (EMP) with operating conditions of 15 kV acceleration voltage, 20 nA beam current, and 1 μm beam diameter. Natural silicate minerals were used as standards, including pyrope (Mg, Al, Si), johannsenite (Ca), rutile (Ti), percivalite (Na), K-feldspar (K), chromite (Fe, Cr), and rhodonite (Mn). Spectral lines, peak counting time (seconds), and off-peak background time (seconds) used for the wavelength-dispersive spectroscopy analyses were as follows: K (Kα, 10, 5), Na (Kα, 10, 5), Cr (Kα, 10, 5), Si (Kα, 10, 5), Al (Kα, 10, 5), Mg (Kα, 10, 5), Ca (Kα, 10, 5), Ti (Kα, 10, 5), Mn (Kα, 10, 5), and Fe (Kα, 20, 10). Detection limits for each of the elements are below 0.01 wt%. Each spot analysis was made adjacent to the LA-ICP-MS analysis pit produced during trace element analysis. Wavelength dispersive X-ray (WDX) maps were collected using an accelerating voltage of 20 kV, a beam current of 100 nA, and a counting time of 50 ms/step.

Trace element analysis of garnet was conducted by LA-ICP-MS at Wuhan Sample Solution Analytical Technology Co., Ltd., Wuhan, China. Laser sampling was performed using a GeolasPro laser ablation system that consists of a COMPexPro 102 ArF excimer laser (wavelength of 193 nm and maximum energy of 200 mJ) and a MicroLas optical system. An Agilent 7700e quadrupole ICP-MS instrument was used to acquire ion-signal intensities. Helium was used as the carrier gas (0.6 l/min). Argon was used as the make-up gas and mixed with helium via a T-connector before entering the ICP. The laser fluence, spot size, and frequency were 6J/cm², 60 µm, and 5 Hz, respectively. National Institute of Standards and Technology (NIST) standard reference material (SRM) 610 was used as the external standard for trace element calibration. The accuracy and precision were monitored by repeated analyses of BCR-2G, BHVO-2G, and BIR-1G standards (Table S2). The preferred values of element concentrations for the NIST SRM 610, BCR-2G, BHVO-2G, and BIR-1G reference glasses are from the GeoReM database (http://georem.mpch-mainz.gwdg.de/). Each analysis involved a background acquisition of approximately 20 to 30 s followed by 50 s of data acquisition from the sample. An Excel-based software, ICPMSDataCal, was used to perform off-line selection and integration of background and analyte signals, time-drift correction, and quantitative calibration for trace element analysis (Liu et al. 2008). Signals were screened visually for heterogeneities such as micro-inclusions or zoning. Trace element concentrations were internally standardized by the Si content of garnet measured using EMP.

Results

Major and trace element compositions of the garnets are listed in Supplementary Tables S3 and S4, respectively. Garnets analyzed consist mainly of andradite-grossular solid solutions, with spessartine, pyrope, almandine and uvarovite collectively less than 5%. Garnets from samples LZS-11a, LZS-29, and LZS-5a exhibit similar compositions, ranging from Adr₆₀Grs₃₉ to Adr₇₂Grs₂₇, Adr₄₇Grs₅₁ to Adr₆₇Grs₃₀, and Adr₅₀Grs₄₈ to Adr₆₁Grs₃₇, respectively (Table S3). In contrast, garnets from sample LZS-32 have substantially higher Fe contents, with compositions of Adr₈₄Grs₁₄ to Adr₉₆Grs₄ (Table S3). Notably, in sample LZS-29, T1 garnet (Adr_47-51) displays lower Fe contents than T2 garnet (Adr_58–66). X-ray mapping reveals that the zoned garnet from sample LZS-5a typically features a relatively Fe-rich core (Adr₅₅) and several overgrowth rims (Adr_50-54) with rhythmic variations in Fe, Al and Mg concentrations, while Ca contents remain relatively constant throughout the grain (Fig. 3).

Fig. 3

Backscattered electron image of the zoned garnet with positions of the analysis marked (A). Major and trace element profile of the zoned garnet showing core-rim variations of andradite contents and ratios of La_N/Yb_N, La_N/Sm_N and Gd_N/Yb_N (B); Compositional maps of Fe (C), Al (D), Mg (E) and Ca (F) for the zoned garnet measured by wavelength dispersive X-ray (WDS) on  electron microprobe

Among the samples, garnets from LZS-11a exhibit the highest REE contents (∑REE = 209.0–489.1 ppm). In sample LZS-29, T1 garnet is more enriched in REE (∑REE = 216.5–344.9 ppm) than T2 equivalent (∑REE = 99.3–167.9 ppm). Garnets from LZS-5a show total REE concentrations ranging from 158.4 to 301.0 ppm, with the zoned garnets having relatively REE-depleted cores (∑REE = 188.6–199.8 ppm) and REE-enriched rims (∑REE = 171.6–301.0 ppm). In comparison with garnets from the other samples, LZS-5a garnets exhibit a wider range of REE contents (∑REE = 63.5–474.2 ppm). Chondrite-normalized REE diagrams (McDonough and Sun 1995) reveal broadly comparable HREE patterns characterized by flat to slightly negative slopes for garnets from all four samples. However, the LREE patterns vary significantly among the different samples. Garnets from samples LZS-11a, LZS-29, and LZS-5a display steep positive slopes (Fig. 4A–C). In contrast, those in sample LZS-32 have moderate negative slopes (Fig. 4D). In addition, garnets from sample LZS-32 display negative but variable Eu anomalies, which are not observed in garnets from the other samples (Fig. 4). For all samples, the garnet shows strong positive correlations between La_N/Yb_N or La_N/Sm_N ratios and the andradite contents, with Gd_N/Yb_N ratios being relatively constant (Fig. 5 A–C). The major and trace element profile for the zoned garnet from LZS-5a illustrates coherent variations in andradite contents and La_N/Yb_N and La_N/Sm_N ratios from the core to the rim, while Gd_N/Yb_N ratios remain relatively constant (Fig. 3B). Overall, the garnet REE patterns do not show much variation with regard to sample location or mineral assemblages (endoskarn vs. exoskarn vs. magnetite ore; Figs. 1, 4), but vary significantly with respect to their garnet major element compositions (i.e., andradite contents; Figs. 4, 5).

Fig. 4

Chondrite-normalized rare earth elements (REE) patterns of the garnets from samples LZS-11a (A), LZS-29 (B), LZS-5a (C) and LZS-32 (D). Garnets from samples LZS–11rare earth elementsa, LZS–29, and LZS-5a display concave shapes (positive slopes of LREE and negative slopes of HREE; A–C), whereas gannets from sample LZS-32 show negative shapes in chondrite-normalized REE diagrams (D)

Fig. 5

Mole fraction of andradite (X_Adr) vs. La_N/Yb_N (A), La_N/Sm_N (B), Gd_N/Yb_N (C) and Eu/Eu^* (D) plots for garnets from the Lazhushan Fe skarn deposit. Eu/Eu^* = Eu_N/ sqrt (Sm_N × Gd_N). Here “N” denotes chondrite-normalized values

Discussion

REE fractionation in garnet

The strong positive correlation observed between La_N/Yb_N ratios and andradite contents in the Lazhushan garnet (Fig. 5A) indicates that the degree of LREE/HREE fractionation in garnet is predominantly linked to its major element compositions. Sush a correlation has been observed for garnets from various skarn deposits (e.g., Gaspar et al. 2008; Xiao et al. 2018; Fei et al. 2019; Wang and Duan 2021). To quantitatively elucidate the relationship between garnet chemistry and REE distribution, we apply the LST of Brice (1975) and Blundy and Wood (1994), which has been effectively utilized for quantifying element partitioning between minerals and melts or fluids (Blundy and Wood 2003; van Hinsberg et al. 2010).

Based on LST equation (Eq. 1), the partition coefficient ratios of two specific trivalent REE (D_REEj/D_REEi, where REEj and REEi are the two different REE elements of interest) can be given by:

$$D_{{{\text{REEj}}}} /D_{{{\text{REEi}}}} = \exp \left\{ {\left( { - 4\pi E_{{{\text{M}}(3 + )}} N_{A} /{\text{RT}}} \right) \times \left[ {r_{0(3 + )} \left( {r_{{{\text{REEi}}}}^{2} - r_{{{\text{REEj}}}}^{2} } \right)/2 \, - \, \left( {r_{{{\text{REEi}}}}^{3} - \, r_{{{\text{REEj}}}}^{3} } \right)/3} \right]} \right\},$$

(2)

As illustrated by Eq. 2, the partition coefficient ratios of D_REEj/D_REEi are dependent on the LST parameters of r₀₍₃₊₎ and E_M(3+) but are independent of D₀₍₃₊₎. To calculate r₀₍₃₊₎ and E_M(3+), we assume that the equations from van Westrenen and Draper (2007), originally calibrated for magmatic garnet, can be applied to hydrothermal garnet. This assumption is based on the consideration that r₀₍₃₊₎ and E_M(3+) are dominantly controlled by the crystal chemistry rather than other factors, such as the melt/fluid compositions (Blundy and Wood 1994; van Hinsberg et al. 2010). According to van Westrenen and Draper (2007), r₀₍₃₊₎ and E_M(3+) can be estimated using the following expressions:

$$r_{0(3 + )} \left( {\mathop {\text{A}}\limits^{ \circ } } \right) \, = 0.9302X_{{{\text{Prp}}}} + \, 0.993X_{{{\text{Grs}}}} + \, 0.916X_{{{\text{Alm}}}} + \, 0.946 \, X_{{{\text{Sps}}}} + \, 1.05\left( {X_{{{\text{Adr}}}} + \, X_{{{\text{Uv}}}} } \right) - 0.0044\left( { \pm 0.0003} \right) \times (P - 3) \, + \, 0.000058\left( { \pm 0.000007} \right) \times (T - 1818),$$

(3)

and.

$$E_{(M3 + )} \left( {{\text{GPa}}} \right) \, = \, 2826\left( { \pm 174} \right) \times \left( {1.38 + r_{0} } \right)^{ - 3} + \, 12.4\left( { \pm 1.3} \right) \times P - 0.072\left( { \pm 0.007} \right) \times T + \, 237\left( { \pm 8} \right) \times \left( {{\text{Al }} + {\text{ Cr}}} \right)_{{{\text{apfu}}}} ,$$

(4)

where P is pressure in GPa, T is temperature in K, (Al + Cr)_apfu denotes the number of atoms per formula unit (12 oxygens) of Al and Cr in garnet, and X refers to the mole fraction.

We use the D_La/D_Yb ratio as an index for the effect of the crystal chemistry on garnet LREE/HREE fractionation. For pure grossular-andradite solid solutions relevant to skarn garnet, X_Py, X_Alm, X_Sps, and X_Uv can be considered approximately 0, whereas X_Grs and (Al + Cr)_apfu approximate to 1- X_Adr and 2(1–X_Adr), respectively. Using the eightfold effective ionic radius of La³⁺ and Yb³⁺ from Shannon (1976), the D_La/D_Yb ratio is calculated by the following equation:

$$D_{La} /D_{Yb} = \exp \left\{ {\left( { - 4\pi E_{M(3 + )} N_{A} /{\text{RT}}} \right) \times [r_{0(3 + )} \times \left( {0.985^{2} - \, 1.16^{2} } \right) \, \times \, 10^{ - 20} /2 \, - \, \left( {0.985^{3} - \, 1.16^{3} } \right) \times 10^{ - 30} /3]} \right\},$$

(5)

where r₀₍₃₊₎ and E_M(3+) are calculated by:

$$r_{0(3 + )} \left( {\mathop A\limits^{0} } \right) \, = \, 0.993 \times \left( {1 - \, X_{{{\text{Adr}}}} } \right) \, + \, 1.05 \times X_{{{\text{Adr}}}} - 0.0044 \times (P - 3) \, + \, 0.000058 \times (T - 1818),$$

(6)

$$E_{{M\left( {3 + } \right)}} \left( {{\text{GPa}}} \right) \, = \,\,2826 \times \left[ {1.38 + \, 0.993 \times \left( {1 - \, X_{{{\text{Adr}}}} } \right) \, + \, 1.05 \times X_{{{\text{Adr}}}} - 0.0044 \times (P - 3) \, + \, 0.000058 \times (T - 1818)} \right]^{ - 3} + \, 12.4 \times P - 0.072 \times T \, + \, 237 \times 2 \times \left( {1 - \, X_{{{\text{Adr}}}} } \right),$$

(7)

According to Eqs. 5,and 6, and 7, the D_La/D_Yb ratio is a function of temperature, pressure, and the andradite contents of garnet. Therefore, under certain P–T conditions, the D_La/D_Yb ratio depends solely on X_Adr. Given a reasonable P–T range of 0.1 to 0.5 GPa and 400 to 600 °C for skarn garnet formation (Meinert et al. 2005), we are able to establish relationships between D_La/D_Yb and X_Adr, which are illustrated in Fig. 6. It is obvious that a pressure variation from 0.1 to 0.5 GPa has virtually no effect on D_La/D_Yb while a variation in the temperature from 400 to 600 °C shows some effect (less than 3 orders of magnitude difference; Fig. 6). Instead, the D_La/D_Yb ratio increases dramatically (up to five to seven orders of magnitude difference) with increasing andradite content (Fig. 6). The modeling results demonstrate the prominent role of garnet composition on LREE/HREE fractionation. This finding explains the strong correlation between the La_N/Yb_N ratios and X_Adr contents observed for garnets from the Lazhushan iron skarn deposit (Fig. 5A). The La_N/Yb_N ratio of garnet is positively correlated with D_La/D_Yb as illustrated by Eq. 8:

$$({\text{La}}_{N} /{\text{Yb}}_{N} )_{{{\text{Grt}}}} = D_{{{\text{La}}}} /D_{{{\text{Yb}}}} \times ({\text{La}}_{N} /{\text{Yb}}_{N} )_{{{\text{fluid}}}} ,$$

(8)

Fig. 6

Correlations between X_Adr and garnet-fluid partition coefficient ratios of D_La/D_Yb (A), D_La/D_Sm (B), and D_Gd/D_Yb (C) based on the modeling using LST. Given that REE reside exclusively in the dodecahedral X site in garnet structure (van Westrenen et al. 1999 and references therein), the eightfold effective ionic radius of REE³⁺ (from Shannon 1976) were used for the LST equations

where (La_N/Yb_N)_Grt and (La_N/Yb_N)_fluid are the La_N/Yb_N ratios of the garnet and fluid, respectively.

The partition coefficient ratios for D_La/D_Sm and D_Gd/D_Yb are used as indices for charting the effect of crystal chemistry on the fractionation of garnet LREE and HREE, respectively. The D_La/D_Sm and D_Gd/D_Yb ratios can be correlated with the andradite content in a manner similar to that for D_La/D_Yb via the equations:

$$D_{{{\text{La}}}} /D_{{{\text{Sm}}}} = {\text{exp}}\left\{ {( - {4}\pi E_{{{\text{M}}({3} + )}} N_{{\text{A}}} /{\text{RT}}) \times [r_{{0({3} + )}} \times \, \left( {{1}.0{79}^{{2}} - { 1}.{16}^{{2}} } \right) \times {1}0^{{ - {2}0}} /{2 } - \, \left( {{1}.0{79}^{{3}} - { 1}.{16}^{{3}} } \right) \times {1}0^{{ - {3}0}} /{3}]} \right\},$$

(9)

$$D_{{{\text{Gd}}}} /D_{{{\text{Yb}}}} = \exp \left\{ {( - 4\pi E_{M(3 + )} N_{A} /{\text{RT}}) \times [r_{0(3 + )} \times \left( {0.985^{2} {-} \, 1.053^{2} } \right) \, \times \, 10^{ - 20} /2 \, - \, \left( {0.985^{3} - \, 1.053^{3} } \right) \times 10^{ - 30} /3]} \right\},$$

(10)

where r₀₍₃₊₎ and E_M(3+) are calculated by Eqs. 6 and 7, respectively. These correlations reveal a significant dependence on the D_La/D_Sm ratios and a weak dependence on the D_Gd/D_Yb ratios with respect to X_Adr (Fig. 6), indicating that LREE is more sensitive to the change of garnet major element composition. It is worth noting that some skarn garnets, predominantly found in W and Sn skarns, can contain appreciable amounts of pyrope (Mg₃Al₂Si₃O₁₂), almandine (Fe₃Al₂Si₃O₁₂) and spessartine (Mn₃Al₂Si₃O₁₂) (Meinert 1992). Incorporating these components into calcic garnet enhances the stiffness of the garnet lattice site (i.e., increasing E_M(3+); Eqs. 3 and 4), leading to steeper correlations than those depicted in Fig. 6.

Compilation of geochemical analyses (n = 1127) of garnet from 22 skarn deposits worldwide provides supporting evidence for the key controls of garnet composition on REE fractionation (Fig. 7; Table S5). Most of the garnets from the dataset are grossular-andradite in composition, with 1098 analyses showing X_Adr + X_Grs > 0.9 (Table S5). It is evident that the garnet La_N/Yb_N ratios from all these skarn deposits are overall strongly correlated with the X_Adr, with rare exceptions where low X_Adr may display high La_N/Yb_N ratios (Fig. 7 A). This observation suggests that, as a first approximation, LREE/HREE fractionation in skarn garnet is primarily controlled by the major element compositions (i.e., X_Adr), despite variations in P, T, fluid composition, and kinetic effects associated with garnet precipitation among those deposits. The apparent positive correlation between La_N/Sm_N and X_Adr and the lack of such a correlation between Gd_N/Yb_N and X_Adr (Fig. 7B, C), aligns with prediction from LST. This contrast further confirms that garnet compositional control is more important for LREE relative to HREE.

Fig. 7

X_Adr vs. La_N/Yb_N (A), La_N/Sm_N (B), Gd_N/Yb (C) and Eu/Eu^* (D) plots for garnets from skarn deposits worldwide

Garnet REE patterns

At the Lazhushan Fe skarn deposit, as the andradite content increases, the garnet REE patterns change from concave shapes (positive slopes of LREE and negative slopes of HREE) to negative shapes in chondrite-normalized REE diagrams (Fig. 4 A––D). Under equilibrium conditions, REE patterns of garnet are determined by garnet-fluid REE partition coefficients and fluid REE patterns. Using D_La as the reference value, the garnet (andradite-grossular)-fluid REE partition coefficient of a specific REE element (REEj) can be calculated via Eq. 11:

$$D_{{{\text{REEj}}}} = D_{{{\text{La}}}} \times \exp \left\{ {\left( { - 4\pi E_{{{\text{M}}(3 + )}} N_{A} /{\text{RT}}} \right) \times \left[ {r_{0(3 + )} \left( {r_{{{\text{La}}}}^{2} - \, r_{{{\text{REEj}}}}^{2} } \right)/2 \, - \left( {r_{{{\text{La}}}}^{3} - \, r_{{{\text{REEj}}}}^{3} } \right)/3} \right]} \right\},$$

(11)

where r₀₍₃₊₎ and E_M(3+) are calculated via Eqs. 6 and 7, respectively. D_La is a constant under certain P, T, fluid chemistry and garnet composition conditions.

The D_La normalized values of garnet-fluid REE partition coefficients are illustrated in Fig. 8. As X_Adr increases, the shape of garnet-fluid REE partition coefficients changes from steep to shallow positive slope (Fig. 8 A, B). To demonstrate the impact of garnet composition on its REE patterns, we assume that the REE pattern of the diorite-derived fluid remains unchanged during garnet formation in the Lazhushan deposit. The REE composition of the diorite-derived fluid is modeled by using the average REE composition of endoskarn garnets (Sample LZS-11a) and the garnet-fluid REE coefficients at 600 °C, 0.1 GPa, and X_Adr = 0.67 (the average X_Adr of Sample LZS-11a). As shown in Fig. 8 D, the modeled fluid has a pattern characterized by strong LREE enrichment and HREE depletion, which is broadly comparable to the whole-rock REE patterns of diorite intrusions in the Daye district where the Lazhushan deposit is located (Li et al. 2009). The pattern of the modeled fluid is also similar to that of granitoid-derived fluid reported elsewhere (e.g., Banks et al. 1994).

Fig. 8

The D_La normalized values of garnet-fluid REE coefficients at 400°C and 0.1 GPa (A) and 600°C and 0.1 GPa (B). The modeled garnet REE pattern at 400°C and 0.1 GPa (C). The modeled REE pattern of garnet and fluid at 600°C and 0.1 GPa (D). Note that Ce and Eu each has two valence states under natural conditions (Ce⁴⁺ and Ce³⁺; Eu²⁺ and Eu³⁺) and the LST equations used in here only constrain the garnet-fluid partition coefficients of Ce³⁺ and Eu³⁺. The eightfold effective ionic radius of REE³⁺ from Shannon (1976)

The garnet REE patterns are modeled using the calculated garnet-fluid REE coefficients and fluid REE composition (Fig. 8C, D). As X_Adr increases from 0 to 1, the REE pattern of garnet gradually changes from HREE-enriched to LREE-enriched shape, with the latter resembling the REE pattern of the modeled fluid (Fig. 8C, D). This phenomenon is because REE fractionation is less effective in garnet with high X_Adr values (e.g., X_Adr > 0.8) (Fig.8 A, B), and fluid REE pattern (LREE-enriched) can overwhelm the influence of garnet-fluid REE partition coefficients. Garnets from the exoskarn sample (LZS-29) and the disseminated magnetite ore sample (LZS-5a) at Lazhushan have X_Adr values ranging from 0.47 to 0.66. These garnets display positive slopes of LREE and flat to moderate negative slopes of HREE (Fig. 4B, C), which match the modeled garnet REE patterns in the same range of X_Adr (Fig. 9A, B). Garnets from the massive magnetite ore sample (LZS-32) have X_Adr values ranging from 0.84 to 0.96 and exhibit LREE-enriched and HREE-depleted patterns (Fig. 4D), broadly consistent with the modeled REE patterns of garnet in the same compositional range (Fig. 9C, D). Although grossular-rich garnets (e.g., X_Grs > 0.8) were not recognized in this study, they, if present, would likely display HREE-enriched patterns (Fig. 8C, D). Overall, the variations in REE patterns of garnet from the Lazhushan skarn deposit can be largely attributed to changing garnet major element composition.

Fig. 9

The modeled REE pattern for the garnets from the exoskarn sample (LZS-29) and disseminated magnetite ore sample (LZS-5a) at 400°C and 0.1 GPa (A) and 600°C and 0.1 GPa (B). The modeled REE pattern for the garnets from the massive magnetite ore sample (LZS-5a) at 400°C and 0.1 GPa (C) and 600°C and 0.1 GPa (D)

The REE patterns of garnet from 22 skarn deposits worldwide are shown in Fig. 10. As garnet composition changes from grossularitic to andraditic, the REE patterns of garnet vary from positive through concave to negative shapes (Fig. 10). This is consistent with the trend observed in the Lazhushan deposit, confirming the significant role of garnet crystal chemistry in shaping garnet REE patterns.

Fig. 10

The Chondrite-normalized REE patterns for garnets from 22 skarn deposits worldwide. Skarn garnets with low andradite contents (X_Adr < 0.25) mostly display positive REE shapes with negative Eu anomalies (A); Skarn garnets with intermediate andradite contents (0.25 < X_Adr < 0.75) commonly display concave REE shapes (B–C). Skarn garnets with high andradite contents (0.75 < X_Adr) generally display negative REE shapes (D). The grey lines represent the overall REE shapes of the given X_Adr range. Garnet REE patterns with zigzag segments and unrealistic anomalies (e.g., strong Tm positive/negative) that may represent incorrect analyses were removed from the figures

The shift in REE patterns, transitioning from HREE-enriched to LREE-enriched shapes, has previously been interpreted as indicative of a transition from equilibrium to kinetic growth of skarn garnet (e.g., Park et al. 2017). REE patterns controlled by the kinetic effect (such as surface sorption) would only be preserved when the rate of mineral growth exceeds the rate of diffusive re-equilibration in the surface layer (McIntire 1963; Smith et al. 2004). Although oscillatory zonation, frequently observed in skarn garnets, is suggestive of such rapid growth, it is not conclusive evidence. In fact, garnets in sample LZS-5a exhibit oscillatory-zoned rims with REE patterns similar to those of the homogeneous core in the same garnet grain (Fig. 4C). The oscillatory zonation in our garnet likely resulted from periodic fluid flux derived from magmas. As demonstrated earlier, the change in REE patterns can be readily explained by an increase in andradite content in garnets that have reached equilibrium with LREE-enriched fluids (Fig. 8). This aligns with the general assumption of local mineral-fluid equilibrium, or its near-attainment, in skarn formation processes (e.g., Reed 1997).

Garnet Eu anomaly

In chondrite-normalized diagrams, skarn garnets often exhibit a decoupling of Eu from its neighboring REE, Sm, and Gd (Figs. 4D, 10). Unlike Sm and Gd, which have a single valence state (Sm³⁺ and Gd³⁺) in natural fluids, Eu exists in two valence states (Eu²⁺ and Eu³⁺). The degree of Eu enrichment or depletion in garnet is quantified by the Eu anomaly, expressed as:

$${\text{Eu}}/{\text{Eu}}_{{{\text{Grt}}}}^{*} = {\text{Eu}}_{{{\text{Grt}} - N}} /{\text{sqrt }}({\text{Sm}}_{{{\text{Grt}} - N}} \times {\text{Gd}}_{{{\text{Grt}} - N}} ),$$

(12)

where Eu/Eu*_Grt is garnet Eu anomaly and the subscript _Grt-N designates chondrite-normalized concentration in garnet. Eu_Grt-N, Sm_Grt-N and Gd_Grt-N can be calculated as follows:

$${\text{Eu}}_{{{\text{Grt}} - N}} = {\text{ Eu}}_{{{\text{fluid}} - N}} \times D\left( {{\text{Eu}}} \right),$$

(13)

$${\text{Sm}}_{{{\text{Grt}} - N}} = {\text{ Sm}}_{{{\text{fluid}} - N}} \times D\left( {{\text{Sm}}^{{{3} + }} } \right),$$

(14)

$${\text{Gd}}_{{{\text{Grt}} - N}} = {\text{ Gd}}_{{{\text{fluid}} - N}} \times D\left( {{\text{Gd}}^{{{3} + }} } \right),$$

(15)

Here, D donates to garnet-fluid partition coefficient and the subscript _fluid-N designates chondrite-normalized concentration in fluid coexisting with garnet. By introducing separate partition coefficients for Eu²⁺ and Eu³⁺, the bulk garnet-fluid partition coefficient for Eu can be calculated as:

$$D\left( {{\text{Eu}}} \right) \, = \, f \times D\left( {{\text{Eu}}^{{{2} + }} } \right) \, + \, \left( {{1} - f} \right) \times D\left( {{\text{Eu}}^{{{3} + }} } \right),$$

(16)

where f is the fraction of Eu²⁺ to total Eu in fluid (f = Eu²⁺_fluid-N/Eu_fluid-N, 0 < f < 1). Considering fluid Eu anomaly:

$${\text{Eu}}/{\text{Eu}}*_{{{\text{fluid}}}} = {\text{ Eu}}_{{{\text{fluid}} - N}} /{\text{sqrt }}({\text{Sm}}_{{{\text{fluid}} - N}} \times {\text{Gd}}_{{{\text{fluid}} - N}} ),$$

(17)

Equations 12–17 can be combined to give:

$${\text{Eu/Eu*Grt }} = {\text{ Eu/Eu*fluid }} \times \, \left\{ {f \, \times \, \frac{{D({\text{Eu}}2 + )}}{{{\text{sqrt}}[D({\text{Sm}}3 + ) \times D({\text{Gd}}3 + )]}} + \, (1 - f) \, \times \, \frac{{D({\text{Eu}}3 + )}}{{{\text{sqrt}}[D({\text{Sm}}3 + ) \times D({\text{Gd}}3 + )]}} \, } \right\},$$

(18)

At 400 to 600 °C, 0.1GPa, D(Eu³⁺) multiplied by the square root of D(Sm³⁺) × D(Gd³⁺) closely approximates 1 (within a range of 1.01 to 1.09) for grossular-andradite garnet (Fig. 11A). To simplify, we use D(Eu³⁺) to replace sqrt[D(Sm³⁺) × D(Gd³⁺)] and this substitution, introducing an error less than 9% for the calculating of Eu/Eu*_Grt. Thus, Eq. 18 can be reduced to:

$${\text{Eu}}/{\text{Eu}}*_{{{\text{Grt}}}} = {\text{ Eu}}/{\text{Eu}}*_{{{\text{fluid}}}} \times \{ {\text{f }} \times [D\left( {{\text{Eu}}^{{{2} + }} } \right)/D\left( {{\text{Eu}}^{{{3} + }} } \right) - {1}] + {1}\} ,$$

(19)

Fig. 11

X_Adr vs. D(Eu³⁺)/sqrt[D(Sm³⁺) × D(Gd³⁺)] (A) and D₀₍₂₊₎/D₀₍₃₊₎ normalized D(Eu²⁺)/D(Eu³⁺) (B) plots. The ratio of D(Eu³⁺)/sqrt[D(Sm³⁺) × D(Gd³⁺)] is independent on D₀ and is calculated with r₀₍₃₊₎ and E_M(3+) derived from Eqs. 6 and 7, respectively, and eightfold effective ionic radius of Eu³⁺, Sm³⁺ and Gd³⁺ from Shannon (1976)

Theoretical predictions from Sverjensky (1984) showed an increase in the stability of Eu²⁺ over Eu³⁺ with increasing temperature and suggested that the Eu²⁺ is prevalent oxidation state for Eu in most crustal fluids at temperatures exceeding 250 °C. A study by Liu et al. (2016) using X-ray Absorption Spectroscopy (XAS) corroborates Sverjensky’s predictions, demonstrating that even under relatively oxidizing conditions (involving a hematite-magnetite-pyrite assemblage) and at pH levels buffered by K-feldspar-muscovite-quartz, the fraction of divalent Eu to total Eu (referred to as “f”) remains above 0.9 at 250 °C. Skarn garnets typically form at temperatures well above 250 °C, in chloride-rich and less oxidized fluids (e.g., Smith et al. 2004; Meinert et al. 2005; Gasper et al. 2008; Xu et al. 2020). This often leads to “f” approaching unity for garnet-forming fluids. When “f” is sufficiently close to unity, Eq. 19 can be simplified as:

$${\text{Eu}}/{\text{Eu}}*_{{{\text{Grt}}}} = {\text{ Eu}}/{\text{Eu}}*_{{{\text{fluid}}}} \times [D\left( {{\text{Eu}}^{{{2} + }} } \right)/D\left( {{\text{Eu}}^{{{3} + }} } \right)],$$

(20)

According to Eq. 20, garnet Eu anomaly is a function of fluid Eu anomaly and D(Eu²⁺)/D(Eu³⁺), the latter of which can be calculated using the following equation:

$$D({\text{Eu}}^{2 + } )/D({\text{Eu}}^{3 + } ) \, = \, \frac{{D_{0} (2 + ) * \exp \{ ( - 4\pi E_{{{\rm M}(2 + )}} {\rm N}_{\rm A} /R{\rm T}) * [r_{0(2 + )} (r_{{{\rm E}u2}} + - r_{0(2 + )} )^{2} /2 + (r_{{{\text{{\rm E}u}}2}} + - r_{0(2 + )} )^{3} /3]\} }}{{D_{0} (3 + ) * \exp \{ ( - 4\pi E_{{{\rm M}(3 + )}} {\rm N}_{\rm A} /R{\rm T}) * [r_{0(3 + )} (r_{{{\rm E}u3}} + - r_{0(3 + )} )^{2} /2 + (r_{{{\text{{\rm E}u}}2}} + - r_{0(3 + )} )^{3} /3]\} }},$$

(21)

where r₀₍₂₊₎ = r₀₍₃₊₎ + 0.04 Å and E_M(2+) = 2/3 E_M(3+) (van Westrenen et al. 2001); r_Eu2+  = 1.25 Å and r_Eu3+  = 1.066 Å (Shannon 1976); r₀₍₃₊₎ and E_M(3+) are calculated via Eqs. 6 and 7, respectively. The parameter of D₀₍₂₊₎/D₀₍₃₊₎ needed in calculating D(Eu²⁺)/D(Eu³⁺) is largely unknown due to the absence of experimental data for garnet-fluid partitioning. This parameter is likely influenced by fluid chemistry due to differences in substitution mechanisms (e.g., Ca²⁺ ⟷ Eu²⁺ verse 2Ca²⁺ ⟷ Eu³⁺  + Na⁺) and speciation effects between divalent and trivalent cations (Brugger et al. 2000; 2008; Smith et al. 2004; van Hinsberg et al. 2010). Therefore, absolute values of D(Eu²⁺)/D(Eu³⁺) cannot be determined in this study. Nevertheless, when normalized to D₀₍₂₊₎/D₀₍₃₊₎, the apparent values of D(Eu²⁺)/D(Eu³⁺) is very low for grossular-rich garnet but increases significantly with increasing X_Adr and temperature (Fig. 11B). This may explain the commonly observed negative Eu anomalies in grossular-rich garnet (Fig. 10A, B) and also the general positive correlation between garnet Eu anomaly and X_adr for skarn garnets in the global dataset (Fig. 7D). Note that, the correlation coefficient (R² = 0.15) is relatively modest for the garnets in the dataset (Fig. 7D), and a negative correlation is observed for garnets in the Lazhushan deposits (mainly caused by Sample LZS-32; Fig. 5D). This suggests that aside from garnet composition, other factors such as fluid Eu anomaly, fluid chemistry and temperature play important roles in controlling garnet Eu anomaly. As indicated by closed-system fractionation modeling, fluid Eu anomaly can vary by several orders of magnitude via precipitating reasonable amounts of scheelite and garnet itself (Brugger et al. 2000; Smith et al. 2004; Xu et al. 2020). Such variation in fluid Eu anomaly would translates into variation in garnet Eu anomaly if D(Eu²⁺)/D(Eu³⁺) remains relatively constant (Eq. 20).

Implications

This study has substantial implications for understanding the REE systematics of skarn garnet. Previous studies have established that variations in the REE–pattern of skarn garnet are primarily associated with changing physiochemical parameters such as temperature, pH, and fluid chemistry as well as mineral growth rate (e.g., Jamtveit and Hervig 1994; Smith et al. 2004; Gasper et al. 2008; Zhai et al. 2014; Xu et al. 2020). These relationships have been widely used to elucidate the origin and evolution of both garnet and skarn-forming fluids (e.g., Park et al. 2017; Zhao et al. 2021; Huang et al. 2022; Ruan et al. 2022; Ghosh and Upadhyay 2022). In this study, we demonstrate a strong dependence of garnet REE patterns on garnet major element composition. Considering the strong correlation between the garnet major element composition and REE fractionations (Figs. 5, 7) and the fact that the garnet composition can change substantially within the scale of the deposit (from grossularitic to andraditic; e.g., Chang and Meinert 2008; Park et al. 2017), it becomes essential to carefully consider the influence of garnet crystal chemistry when interpreting garnet REE patterns.

Our study further illustrates that the LST equations allow for the prediction of fluid REE patterns based on garnet REE patterns, and vice versa. We demonstrate that the Eu anomaly in skarn garnet is mainly a function of fluid Eu anomaly and the garnet-fluid partition ratio of D(Eu²⁺)/D(Eu³⁺). Garnet composition exerts controls on D(Eu²⁺)/D(Eu³⁺) and consequently affects garnet Eu anomaly. These findings underscore the crucial role of garnet crystal chemistry in the REE fractionation of skarn garnets. Taking this factor into account can significantly enhance the reliability of using garnet REE as tracers in skarn systems.